The purpose of this study is the computer analysis of digitized images of urothelial cells from the urinary sediment. The digitized images are obtained by scanning cells stained with Papanicolaou method by means of a Zeiss SMP scanning microscope at 0.5 intervals, on line with PDP-12 (36K) computer. The cell images are forwarded on tape to the University of Chicago for analysis on dedicated PDP-10 computer wherein resides the Taxonomic Intracellular Analytic System (TICAS) for cell analysis. In the initial stage of this project, a computer analysis of benign (NEG) and malignant (POS) cells could be carried out with a high degree of accuracy (plus or minus 10%), by the use of both the supervised learning program DISTILL and unsupervised learning algorithm PINDEX. The next step in this program has been computer classification of cells impossible to classify by visual means, hence atypical (ATY). For each one of the 21 features tested, the values for ATY cells were intermediate to the values of NEG and POS cells. Using the PINDEX program, the ATY cells could be split into subsets (NEG, POS and ATY) and allowed the construction of individual patient charts. During the year 1976-1977 a clinical trial of the TICAS system for the diagnosis of bladder cancer was conducted. The computer-generated diagnoses on 8 patients appear very promising. However, a number of technical problems must still be solved before the system can be subjected to field trials.